Response error in a web survey and a mailed questionnaire: The role of cognitive functioning

Web-based interviewing is gradually replacing traditional modes of data collection, in particular telephone and mailed surveys. This global trend takes place despite the fact that established knowledge of its consequences on response error is incomplete. This paper studies differences between a web (CAWI) and a mailed version (MAIL) of a questionnaire in various forms of response error, namely item nonresponse, satisficing, person-reliability, and social desirable responding. We posit 1) that response error depends on respondents cognitive functioning, namely in the domains of global reading abilities, fluid intelligence, as well as working and episodic memory; and 2) that these effects differ across modes of data collection with generally higher prevalence in the CAWI mode since this mode is more demanding. The analysis builds on a randomized mode experiment implemented in the context of the Berlin Aging Study II (BASE-II), a survey that primarily focuses on multidimensional processes of physical and mental aging (see Bertram et al. 2014). The analysis reveals a high impact of cognitive functioning at the various stages of the survey response process. While we do found moderate mode-differences in response error, such as higher item nonresponse rates in the CAWI mode, we did not find cognitive functioning to be a better predictor of response error in web-based interviewing.